The paper investigates to what extent linguistic features, primarily emotional and affective signals, serve as a common denominator in the online texts of various extremist groups. Examining social media data from multiple sources, this research focuses on ideologies such as far-right, jihadist, and various manosphere subgroups. Key results demonstrate distinct emotional profiles across ideologies, with "disgust" emerging as a dominant emotion in many extremist discourses, while "joy" is most prominent in the analyzed religious extremist content. Statistically significant dissimilarities in emotional expression patterns were confirmed, allowing for differentiation between ideological groups. Classifiers, particularly the DeBERTa model, demonstrated a strong capability to reliably distinguish between extremist groups based solely on linguistic characteristics.